NORMATIVITY AND INTENTIONALITY
DRAFT Anandi Hattiangadi
The hard problem of intentionality is to explain what makes it the case that an
arbitrary sentence or thought has the semantic properties that it does rather than
some other semantic properties or none at all. Some hold that intentionality is
normative, and that this has a crucial bearing on the hard problem of intentionality.
This paper investigates whether this is so.
It is possible to distinguish four versions of the thesis that intentionality is
normative: a) grasp of a concept or meaning involves following a rule or making a
normative judgment of some kind; b) the concepts of meaning and content are
normative; c) meaning and content are sources of normativity; d) the semantic facts
are in some sense reducible to normative (and natural) facts. I discuss all four
versions of the thesis, and argue that the normativity of intentionality has little
bearing on the hard problem of intentionality.
1 INTRODUCTION Philosophy, linguistics, and psychology have furnished us with a vastly improved
understanding of intentionality, the capacity to represent something or other in some
1 way. Yet the so-called ‘hard problem’ of intentionality remains unresolved.1 This is
the problem of explaining how and why the semantic properties of intentional
states—their meaning or content2—arise in the world as described by the natural
sciences, such as most notably, physics, chemistry and biology. Since intentionality is
unlike anything else we encounter in the natural world, it gives rise to some hard
questions: can the semantic truths be reductively explained in terms of the natural
truths, or more broadly, in terms of the non-semantic truths? Do the semantic facts
supervene3 on the non-semantic facts? What makes it the case that an arbitrary
representation has the semantic properties that it does rather than some other semantic
properties or none at all?
This essay focuses on approaches to the hard problem of intentionality that in
some way or other appeal to the idea that intentionality is normative. This idea took
shape in Saul Kripke’s influential discussion of Wittgenstein’s rule following
considerations. Recall that Kripke asks what makes it the case that he means addition
by ‘plus’ rather than quaddition (where x quus y = x plus y if x, y ≤ 57, and = 5,
1
Chalmers 1996 distinguishes the hard problem of consciousness, which is to reductively
explain consciousness in non-phenomenal terms, from the easy problems, having to do with
such things as finding correlations between conscious experience and states of the brain, or
with understanding the relationships between different kinds of conscious experience. For
Chalmers, the distinction between the hard problems and the easy problems has to do with the
methods by which they can be resolved: the easy problems can be resolved by ordinary
empirical methods, whereas the hard problems require a priori reasoning. This distinction
between methods appropriate to the resolution of hard and easy problems is not directly
relevant to the present discussion.
2
Typically, meaning is said to be a property of sentences, expressions or utterances, while
content is said to be a property of mental states, such as beliefs, desires, intentions, concepts
or mental representations. It may be that there is a kind of intentionality that does not involve
meaning or content. We can set aside that kind of intentionality here.
3
Roughly, to say that the A-facts supervene on the B-facts is to say that any two things that
are qualitative duplicates in all their B-properties are duplicates in all their A-properties.
There are many varieties of supervenience. For a survey of them, see CITE.
2 otherwise), and reaches the skeptical conclusion that there is no fact of the matter
what any word means. He rejects the view that meaning is determined by speaker
dispositions on the grounds that ‘the relation of meaning and intention to future action
is normative, not descriptive (Kripke 1982: 37).’ The problem, Kripke says, is ‘that
my present mental state does not appear to determine what I ought to do in the future
(Kripke 1982: 56).’ Similarly, in response to the view that meaning is determined by
an experience with a distinctive qualitative feel, he asks, ‘How on earth would this
[experience] help me figure out whether I ought to answer ‘125’ or ‘5’ when asked
about ‘68+57’? If the [experience] indicates that I ought to say ‘125’, would there be
anything about it to refute a sceptic’s contention that, on the contrary, it indicates that
I should say ‘5’ (Kripke 1982: 42)?’ Though Kripke directly discusses the normativity
of meaning, it is clear that his claims must generalize, and he is committed to the
normativity of content as well (Boghossin 1989, Hattiangadi 2007).
Inspired in part by Kripke, several philosophers have argued that meaning is
normative (Boghossian 1989, 2005 Brandom 1994, 2000, Gibbard 1994, 2013,
Ginsborg 2011, 2012, Whiting 2007), content is normative (Boghossian 2003,
Gibbard 1996, 2003, Wedgwood, 2007), or more generally that intentionality is
normative (Wedgwood 2007). One underlying motivation for this is the thought that
there is a useful analogy to be drawn between the hard problem of intentionality and
the hard problem of normativity—the problem of explaining what makes it the case
that something is right, wrong, good, bad, obligatory, forbidden or permitted in nonnormative terms. This certainly seems to be Kripke’s motivation. Though one might
reject Kripke’s anti-realism about normativity, and resist his skeptical conclusion
about meaning, it is worth considering whether the normativity of meaning or content
has any bearing on the hard problem of intentionality. If it does, then well-developed
3 approaches to the hard problem of normativity might be fruitfully carried over to the
hard problem of intentionality.
What exactly does it mean to say that intentionality is normative? First of all,
it will help to distinguish between normativity as a feature of a certain class of
representations and normativity as a feature of the world. It is relatively
uncontroversial that there are normative judgments, statements, concepts and
expressions, such as the statement ‘FGM ought to be prohibited’, the judgment that all
and only pleasure is good, the expression ‘good’ as used in ‘all and only pleasure is
good’ and the concept ‘good’ is used to express. More controversial is the thought
that in addition to normative representations, there are normative facts and properties
to which normative representations correspond, such as the fact that FGM ought to be
prohibited or the property of goodness.
Those who claim that intentionality is
normative can be viewed as using ‘normative’ in either of these two ways. Indeed, we
can distinguish four hypotheses—two of each kind.
1. Meaning or content essentially involves normative judgment. This is the view
that what it is to grasp a concept or a meaning involves following a
constitutive rule, or more generally, that grasping a meaning or concept
involves making a normative judgment of some kind. For instance, a
proponent of this view might say that if Shah means addition by ‘plus’, then
she follows a rule for addition, or is disposed to judge that ‘57+68=125’ is
correct.
2. Semantic concepts are normative. According to this view, the concept
‘meaning’ is normative, as are straightforward ascriptions of meaning and
4 content. For instance, a proponent of this view might say that the statement
‘Shah means addition by “plus”’ is a normative statement and the judgment
that Shah believes that 57+68=125 is a normative judgment.4
3. Meaning and content are sources of normativity. This is the view that
semantic facts or properties determine or give rise to normative facts or
properties. For instance a proponent of this view might say that necessarily, if
Shah means addition by ‘plus’, then she ought to accept that 57+68=125, or at
least that she has a normative reason to do so.5,6
4. Semantic facts are reducible to the normative (and natural) facts. According
to this view, normative facts, perhaps together with the physical facts,
determine the semantic facts. For instance, a proponent of this hypothesis
might say that necessarily, if Shah ought to accept a sentence s iff p is true,
then s means that p in the language Shah speaks.
These four versions of the hypothesis that intentionality is normative are not mutually
exclusive. Indeed, they are often held in tandem. But since they are independent, it
will make sense to consider them separately. Is intentionality normative in any of
these four senses? And what does the normativity of intentionality in any of these four
senses have to do with the hard problem of intentionality? As we shall see,
normativity ultimately has very little bearing on the hard problem of intentionality.
4
Not all such statements and judgments, however. For instance, sometimes statements
involving normative terms are used in the ’inverted commas sense’, that is, as an implicit
quotation.
5
This is what Kathrin Glüer and Åsa Wikforss call ME/CE normativism, for meaningengendered and content-engendered norms.
6
I assume that to accept a sentence is either to assert it, assent to it, or to be disposed to do so.
5 2 MEANING, CONTENT AND NORMATIVE JUDGMENT Kripke introduces the thought that meaning is normative in the context of a discussion
of Wittgenstein’s rule following considerations. At the heart of these reflections on
rule following is the idea that language is a conventional activity, which invites an
analogy between language and other conventional activities such as games
(Wittgenstein 1953): just as the activity of playing bridge is in part constituted by the
rules of bridge, so too is the activity of speaking a language. Kripke suggests that
meaning essentially involves rule following, as when he says that ‘[o]rdinarily, I
suppose that, in computing ‘68+57’ as I do, I do not simply make an unjustified leap
in the dark. I follow directions I previously gave myself that uniquely determine that
in this new instance I should say ‘125’ (1982:CITE),’ and that such directions ‘must
somehow be “contained” in any candidate for the fact as to what I meant
(1982:CITE).’
The hypothesis that speaking a language involves rule following is a special
instance of the more general thesis that meaning something by an expression, or
grasping a concept, essentially involves making a normative judgment of some kind.
As is widely agreed, following a rule involves more than merely acting in accordance
with it; it involves in some sense being guided by or accepting the rule. And this
involves a normative judgment. Even if the idea that meaning involves rule-following
is waning in popularity, many philosophers are drawn to the more general thesis that
meaning something or grasping a concept essentially involves a normative judgment
of some kind. For instance, normative inferentialists, such as Robert Brandom, hold
that for a speaker to grasp the meaning of a sentence she must participate in a practice
by accepting certain inferential rules, such as a rule that permits inference from ‘roses
6 are red’ to ‘roses are coloured’, and ‘entrance’ rules, which specify the conditions
under which one is permitted to assert a sentence, such as when one has suitable,
undefeated perceptual evidence for its truth (Brandom 1994). Similarly, conceptual
role semanticists, such as Christopher Peacocke (1992) and Ralph Wedgwood (2009),
hold that grasping a concept such as ‘if’ involves normative commitments and
normative judgments. For Wedgwood, grasp of the concept ‘if’ involves a rational
disposition to accept inferences by modus ponens: ‘if A then B; A; therefore B’
(Wedgwood 2009). A rational disposition is a disposition that is sensitive to the
rationality of a transition in thought. To grasp the concept ‘if’, it is not enough to be
disposed accept inferences by modus ponens; it must be the rationality of those
inferences that trigger the disposition. Thus, grasp of a concept involves at least an
implicit judgment that certain patterns of inference are rational. Assuming that
judgments of what is rational are normative, it follows that grasp of a concept
involves a normative judgment of some kind (See also Boghossian 2003, Ichikawa
and Jarvis 2013).
Similarly, Hannah Ginsborg defends a view of this general kind, though she
claims that meaning essentially involves implicit rule following rather than explicit
rule following. What it is to implicitly follow a rule for Ginsborg is to be disposed to
make the appropriate pattern of normative judgments. For instance, on her view, it is
necessary for meaning addition by ‘plus’ that a subject be disposed to judge that
‘57+68=125’ is correct, that ‘125’ is the answer one ought to give to ‘68+57=?’. She
describes this as a ‘primitive’ attitude in the sense that it need not be the result of
conscious rule following, or even the product of a post hoc realization of which rule
one has been following (Ginsborg 2011).
7 Does intentionality involve normative judgment, as these philosophers
suggest? There are some excellent reasons to think not. First of all, rules that might be
thought to be constitutive of meaning are not the sorts of rules that can be followed
(Glüer & Pagin 1999, Glüer 1999). Consitutive rules of meaning specify what a
sentence means, and this does not naturally coincide with telling speakers what to do.
Consider, for instance, a rule that says: ‘plus’ means addition in English. This is a
descriptive rule, which captures a generalization about usage. Since it does not tell
anyone what to do, it is not at all obvious what is involved in following it. Indeed,
even if you accept this rule, your acceptance of it can only play the role of a belief in
practical reasoning (Cf. Glüer and Pagin 1999, Glüer & Wikforss 2010).
Some have sought to sidestep this problem by formulating constitutive rules in
terms of correctness conditions, and then arguing that correctness is a normative
notion (Boghossian 1989, 2003, Gibbard 2003). The rule for meaning addition by ‘+’
might then be stated as follows:
(1) ϕ + ψ = ω is correct iff the number denoted by ω is the sum of the
⎡
⎤
numbers denoted by ϕ and ψ.
The crucial claim is that ‘correct’ is a normative term, having to do with what one
ought to do or may do (Boghossian 2003, Gibbard2003). So, to accept (1) is to make a
normative judgment. But ‘correct’ is not a normative term across all contexts (Bykvist
& Hattiangadi 2013), and it is far from obvious what the normative implications of (1)
are supposed to be. If I accept (1), do I judge that I ought to assert every addition fact,
8 or merely that I may do so? Do I judge that I ought to blurt out addition facts at any
time, or only when I am asked? A natural reading of (1) is that in order to mean
addition by ‘plus’ you need to take yourself to be under an obligation to respond with
the sum of any two numbers when asked. But this view is implausible. Suppose that
you discover out that if you answer with the sum of two numbers in some situation,
thousands will die. Though you will undoubtedly judge in this situation that you
ought not to respond with the sum, we can hardly conclude that you do not mean
addition by ‘plus’. Similar considerations tell against the interpretation of (1)
according to which in order to mean addition by ‘plus’ you must judge that you may
respond with the sum of any two numbers when asked (E.g. Whiting 2009). For in the
circumstance just described, you sensibly judge that you ought not to respond with the
sum, which is incompatible with the judgment that you may do so. Nearly as
implausible is the thought that meaning addition by ‘plus’ requires that you judge
yourself to have a pro tanto reason to assert ‘57+68=125’, that is, a reason that can be
outweighed by other considerations. However, an unrepentant liar might not take
there to be any even pro tanto reason to assert an addition truth in some situation.
Once again, we cannot conclude that the unrepentant liar does not mean addition by
‘plus’ (Cf. Glüer 1999, 2001, Hattiangadi 2006, 2007, Wikforss 2001).
Perhaps the conclusion to draw is that correctness is a sui generis normative
notion, irreducible to obligation and permission (See McHugh 2013). One might then
say that the liar must at least judge that responding with the sum is correct in view of
what ‘plus’ means, even if she does not take herself to have any reason to do so. But it
is difficult to see in what sense correctness is a normative notion if it is possible to
coherently judge that some act is correct while simultaneously judging that one has no
normative reason to perform that act. Even if correctness is irreducible to any other
9 normative notions, it must bear some relation to them. McHugh suggests that if one
judges an act to be incorrect, then one judges that one has a pro tanto reason not to
perform that act. But as the case of the unrepentant liar proves, it is possible for
someone to think that they have no pro tanto reason not to respond with the sum in
some situation without thereby failing to mean addition by ‘plus’.
Ginsborg puts a kind of particularist twist on the idea that meaning involves
rule following. According to her, to mean something by an expression, a subject need
not explicitly endorse any rule, but must be disposed to make the appropriate pattern
of particular judgments of correctness. Yet the particularist twist does not really help
to address the foregoing objections. Since judgments of correctness are not normative
in every context, we need to know what the normative implications of these primitive
judgments of correctness are. Perhaps Ginsborg will say that to mean addition by
‘plus’ you must be disposed to judge that you ought to answer with the sum of any
two numbers when asked. But consider once again the situation in which you judge
that you ought not to assert that 57+68=125 in order to avoid the loss of life. Though
you are not disposed to make the primitive judgment that is constitutive of meaning
addition by ‘plus’, it does not follow that you do not mean addition by ‘plus’.
Perhaps Ginsborg will instead say that to judge that some answer is correct is to judge
that you have a normative reason to give that answer. But once again, an unrepentant
liar who sees herself as having no reason to say that 57+68=125 in some contexts,
need not fail to mean addition by ‘plus’.
Perhaps the suggestion that grasping a concept essentially involves a
normative judgment fares better. Consider the conceptual role semanticist’s
suggestion that in order to grasp the concept ‘if’, a subject must have a rational
disposition to accept arguments by modus ponens. We can put pressure on this
10 suggestion by appeal to Timothy Williamson’s example of Vann McGee, a
distinguished logician who has presented counterexamples to modus ponens, such as
the following (Williamson 2007:92ff). Suppose that the opinion polls place the
Republican Ronald Reagan decisively ahead of the Democrat Jimmy Carter, who in
turn is placed decisively ahead of the other Republican, John Anderson. It is
reasonable to believe that if a Republican wins the election, then if it is not Reagan
who wins it will be Anderson. It is also reasonable to believe that a Republican will
win the race. Yet McGee argues that it would not be reasonable to conclude that if it
is not Reagan who wins, it will be Anderson.
In presenting this counterexample, McGee shows that he lacks the rational
disposition to accept arguments by modus ponens in all cases, for in some cases, he
judges it irrational to accept an argument by modus ponens. Yet, as a distinguished
logician, it is unquestionable that McGee grasps the concept of the conditional; if
anything, he is an expert on conditionals (Williamson 2007). So, it appears to be
possible to grasp the concept of the conditional while rejecting as irrational some
arguments by modus ponens.
Perhaps it will be said that though McGee does not accept all arguments by
modus ponens, he accepts enough of them to come close enough to qualify as
competent in his grasp of the concept of the conditional. The trouble with this reply is
that though McGee may come close enough to qualify as competent with the concept
of the conditional, he better qualifies as competent with a distinct concept—call it the
concept of the McGee conditional. Let’s say that to grasp the McGee conditional you
have to be rationally disposed to accept inferences by McModus ponens: ‘if A then B;
A; therefore B, in all contexts except the McGee cases’, where the McGee cases are
those cases that McGee puts forward as counterexamples to modus ponens. McGee
11 better qualifies as grasping the concept of the McGee conditional than as grasping the
conditional, for he accepts all inferences by McModus ponens, and this is constitutive
of the concept of the McGee conditional. If what it is to grasp a concept is to have the
rational dispositions constitutive of that concept, we should say that McGee grasps the
concept of the McGee conditional not the conditional.
Could it be that McGee cannot have a rational disposition to accept arguments
by McModus ponens, because that rule is not a requirement of rationality? If
McModus ponens is not a rational rule, it is not possible to have a disposition to
accept arguments by McModus ponens because those arguments are rational.
However, this reply has the implausible consequence that it is impossible for anyone
to grasp the concept of the McGee conditional. Though McGee himself surely grasps
the concept of the conditional, it is possible that someone with McGee’s dispositions
grasps the concept of the McGee conditional. But if it is a necessary truth that one
rationally ought to accept arguments by modus ponens, then it is not possible for
anyone to have a rational disposition to accept arguments by McModus ponens, and it
is not possible for anyone to grasp the concept of the McGee conditional. But this
consequence is unacceptably strong. Though admittedly bizarre, the concept of the
McGee conditional is a bona fide concept. It is easy enough to grasp the concept
whether or not one accepts that McGee’s cases really constitute counterexamples to
modus ponens.
It is time to move on to the question whether the hypothesis that meaning and
content involve normative judgments of some kind has any bearing on the hard
problem of intentionality. Kripke certainly suggests that it does. He takes the upshot
to be that there are no semantic facts, ‘no such thing as meaning anything by any
word’. Similarly, both Brandom and Ginsborg present themselves as addressing
12 Kripke’s problem and proposing an alternative solution. Perhaps these authors have in
mind a different version of the thesis that intentionality is normative. Because even if
it is true that meaning and content involve normative judgments of some kind, this
ultimately has no bearing on the hard problem of intentionality.
It is important to distinguish the hard problem of intentionality from what
might be called ‘first order’ semantics.7 The principle aim of first order semantics is
to accurately model the intentional and semantic phenomena. Among other things, it
concerns the relations between different kinds of intentional and semantic
phenomena, including relations between the meanings of expressions and the
meanings of sentences, and relations of priority or determination between different
kinds of intentional state. For instance, it is a question of first order semantics whether
sense determines reference, or whether reference is determined by causal-historical
relations to an initial dubbing event. It is easy to confuse this first order determination
relation with meta-semantic determination.8 What makes the causal-historical theory
of reference a first-order theory is that the postulated determination relation holds
between two kinds of intentional phenomena: contemporary uses of a referring
expression and a prior intentional act of naming or stipulation, the ‘dubbing event’. In
the dubbing event, the intentions of the dubber play a crucial role in fixing the
referent of the expression.9
7
This roughly corresponds to what Stalnaker (CITE) calls descriptive semantics, but the use
of the term ’descriptive’ in this context might be misleading.
8
Cf. Stalnaker (CITE). My reason for this is that the causal-historical theory relates current
use of a referring expression to a prior intentional act—the dubbing event. In contrast, a
causal theory of reference in the
9
In contrast, Fodor’s (CITE) and Dretske’s (CITE) theories of mental representation fall
squarely in the meta-semantics. Though these theories also appeal to causal relations, they
aim to be fully reductive.
13 Now, it is clearly a question of first order semantics whether meaning
something by an expression or grasping a concept essentially involves making a
normative judgment of some kind. This is because the question whether meaning or
content involves normative judgment has to do with relations between intentional
mental states, between different kinds of intentional or semantic phenomena—
between the intentional mental state of meaning something by an expression or
grasping a concept and the intentional mental state of following, endorsing or
accepting a rule, or making a normative judgment of correctness. Construed as an
answer to a first-order question concerning the interrelations between semantic
phenomena, the hypothesis that meaning or content involves normative judgment
might well be illuminating. But construed as an answer to the hard problem of
intentionality, it is blatantly circular. The hard problem of intentionality is to
reductively explain intentionality in non-semantic, non-intentional terms, and any
unreduced appeal to normative judgments in the reductive explanation of
intentionality would render the explanation circular. Of course, a dualist with respect
to intentionality will argue that the hard problem of intentionality cannot be solved.
But it is hard to see how the hypothesis that meaning and intentionality involve
normative judgments would speak in favour of dualism. Boghossian (1989) suggests
that the normativity of meaning and content figure in an argument for dualism.
However, this conclusion can only be drawn if some other version of the normativity
thesis is true, such as that semantic requirements are sources of normativity, or if
normative truths figure essentially in the reductive explanation of intentionality.
Some have suggested that the hypothesis that meaning or content involve
normative judgment constitutes a partial solution to the hard problem of intentionality
(Ginsborg 2011). The idea might be that the explanation of intentionality proceeds in
14 stages: first, we reductively explain meaning and content in terms of normative
judgments, and this paves the way for a reductive explanation of the contents of
normative judgments. Ginsborg suggests that the contents of these normative
judgments can be reductively explained in terms of dispositions, and that Kripke’s
objections to the dispositional theory can be met (though she doesn’t herself say
how). But if Kripke’s objections to the dispositional theory can be met, a dispositional
explanation of the contents of non-normative judgments is available and it is possible
to avoid the step of explaining meaning in terms of normative judgments. Thus, to be
told that meaning or content involve rule following is not immediately relevant to the
hard problem of intentionality.
3 THE NORMATIVITY OF SEMANTIC CONCEPTS
In his recent book, Allan Gibbard argues not that meaning is normative, but that the
concept MEANING is normative,10 much like paradigmatic normative concepts such as
‘good’ and ‘ought’. He claims further that meta-representational ascriptions of
meaning, such as ‘Shah means addition by “plus”’ are normative, not descriptive, as
are self-ascriptions of meaning, and ascriptions of propositional attitudes. Gibbard
combines this view with his own brand of expressivism about normativity according
to which assertions of normative statements do not purport to describe how things are,
but express something akin to plans. Gibbard proposes to shed light on the hard
problem of intentionality by exploiting the analogy with meta-ethics. In meta-ethics,
the expressivist says that because moral sentences are non-descriptive, there is no
need to postulate any moral facts out there in the world to correspond to them, and a
10
For the duration of this section, I will adopt Gibbard’s convention of using small capital
letters to denote concepts and thoughts.
15 fortiori, no moral facts crying out for reductive explanation. By the same token, if
semantic ascriptions express plans, there is no need to postulate any semantic facts
out there in the world to correspond to them, and a fortiori, no semantic facts crying
out for reductive explanation.
Gibbard introduces his view with the slogan ‘means entails ought’.
‘Ascriptions of meaning,’ Gibbard says, ‘imply straight ought ascriptions’ (Gibbard
2012: 11). He adds that these entailments ‘could only be analytic or conceptual’
(Gibbard 2012:23). That is, ascriptions of meaning, such as ‘Pierre means
DOG
by
“chien”’, analytically entail ascriptions of oughts, such as ‘Pierre ought to…’; these
entailments hold in virtue of the meaning of ‘meaning’ or equivalently, in virtue of
the concept
MEANING.
Anyone who grasps the concept
MEANING
must accept the
normative entailments of meaning ascriptions.
What are the normative entailments of meaning ascriptions? Here are two
examples drawn from Gibbard’s discussion, where the first statement of each pair is
said to analytically entail the second:11
1a. The sentence ‘schnee ist weiss’ in Ursula’s language means
SNOW IS
WHITE.
11
The first example occurs on pg. 40 of Gibbard 2012, with minor stylistic variations, while
the second is reconstructed from Gibbard’s discussion of analyticity and synonomy in chapter
6. Some further examples occurring in the text are somewhat puzzling. For instance, Gibbard
says: ‘I ought not to believe, all at once, that snow is white and that nothing is white—and
that ties in with the meaning of our term “nothing”’ (Gibbard 2012:13). Even if it is true that I
ought not to believe both that snow is white and that nothing is white, this seems not to have
anything to do with the meaning of the term ‘nothing’. For it is arguably true that Kaveri, a
monolingual speaker of Konkani, ought to not to believe both that snow is white and that
nothing is white, yet this has little to do with the meaning of our term ‘nothing’.
16 1b. Ursula ought to accept ‘schnee ist weiss’ iff snow is white.
2a. Pierre’s sentence ‘les chiens aboient’ means DOGS BARK.
2b. If Pierre has sufficient undefeated evidence that dogs bark in an epistemic
circumstance E, then Pierre ought to accept the sentence ‘les chiens aboient’ in
E.
If these entailments are analytic as Gibbard suggests, then anyone who grasps the
concept
MEANING
must accept that 1a entails 1b, and that 2a entails 2b, on pain of
irrationality or conceptual confusion. However, it is possible for someone to sensibly
deny these entailments without thereby displaying either irrationality or confusion
about the concept MEANING.
First, consider an evidentialist, such as William Clifford, who holds that one
ought never to believe anything on insufficient evidence (Clifford 1877). Suppose that
we present Clifford with a hypothetical scenario in which snow is in fact white;
Ursula is a speaker of German; but Ursula has never seen snow, and lacks any
testimonial evidence as to its colour. Since Ursula lacks sufficient evidence that snow
is white, Clifford would judge that Ursula ought not to believe
Furthermore, according to Gibbard, one way to believe
SNOW IS WHITE.
SNOW IS WHITE
is to accept a
sentence in a language one understands that means that snow is white (Cf. Gibbard
2012:27, 2012:44). So, Clifford’s judgment that Ursula ought not to believe
WHITE
SNOW IS
is tantamount to the judgment that Ursula ought not to accept the sentence
‘schnee ist weiss’. Thus, Clifford would accept 1a but not 1b. Yet, in so doing, he
17 would display neither irrationality nor confusion about the concept MEANING. Even if
we think Clifford is mistaken, his mistake does not result from a failure to grasp the
concept MEANING, because his reasons for rejecting the entailment from 1a to 1b have
to do with his acceptance of evidentialism, a substantive view in epistemology. Thus,
1a does not analytically entail 1b (Cf. Whiting 2015).
It can be shown in a similar fashion that the entailment from 2a to 2b is not
analytic. Consider Blaise Pascal (1670), who holds that there can be pragmatic
reasons for belief. Suppose that we present Pascal with a hypothetical situation in
which Pierre is a monolingual speaker of French who has sufficient undefeated
evidence that dogs bark, but is promised eternal bliss if he does not accept the
sentence ‘les chiens aboient’ and eternal torture if he does. With regard to this
situation, Pascal would accept 2a, but deny that Pierre ought to believe
Since Gibbard assumes that believing
DOGS BARK
DOGS BARK.
is tantamount to accepting a
sentence in one’s language that means DOGS BARK, it follows that Pascal would reject
2b. Yet in so doing, he need neither display irrationality nor confusion about the
concept MEANING. Even if Pascal’s view that there can be pragmatic reasons for belief
is mistaken, his mistake is not conceptual. His grounds for rejecting the inference
from 2a to 2b have to do with his pragmatism, not with his grasp of the concept
12
MEANING.
Thus, 2a does not analytically entail 2b.
One way to respond to these objections might be to appeal to the notion that
there are distinct spheres of normativity: such as inter alia, epistemic, prudential and
12
One might try to argue that he fails to grasp the concept of belief, because beliefs cannot be
formed at will, but that would not help Gibbard in his efforts to argue that the concept of
meaning is normative. In any case, Gibbard accepts that beliefs can be formed at will
(Gibbard 2012).
18 semantic normativity. One might then say that what 1a and 2a entail are 1c and 2c,
respectively:
1c. Ursula semantically ought to accept ‘schnee ist weiss’ iff snow is white.
2c. If Pierre has sufficient undefeated evidence that dogs bark in an epistemic
circumstance E, then Pierre semantically ought to accept the sentence ‘les
chiens aboient’ in E.
One could argue that even if Clifford could sensibly reject the inference from 1a to
1b, he would have to be conceptually confused to reject the inference from 1a to 1c.
What Clifford ought to say about Ursula’s case is that she epistemically ought not to
believe that snow is not white, but that she semantically ought to accept the sentence
‘schnee ist weiss’ nonetheless. Similarly, one could argue that even if Pascal could
sensibly reject the inference from 2a to 2b, he would have to be conceptually
confused to reject the inference from 2a to 2c. What Pascal should say is that while
Pierre prudentially ought not to believe
DOGS BARK,
he semantically ought to accept
the sentence ‘les chiens aboient’ nonetheless.
However, this response is not available to Gibbard, who claims that semantic
ascriptions entail ‘Ewing’s primitive oughts’, where to say ‘you ought to do X’ is to
say that you ought to do X all things considered, or that you have a conclusive reason
to do X (Gibbard 2012:14). Thus, Gibbard’s official view is that 1a entails 1b, and 2a
entails 2b, where both 1b and 2b are understood as all things considered oughts, not
that 1a entails 1c nor that 2a entails 2c.
19 Moreover, dropping the suggestion that semantic oughts are Ewing oughts
does not help. It is possible for someone to sensibly reject the inference from 1a to 1c,
or from 2a to 2c, without displaying any conceptual confusion, so even these
entailments are not plausibly analytic. To see why, it will help to generalize 1c and
2c. Let S be a subject; assume that sentence ‘s’ means P in the language spoken by S,
and that ‘s’ is true iff p. We can generalize 1c and 2c as follows:
1d. S semantically ought to accept ‘s’ iff p.
2d. If S has sufficient undefeated evidence that p is true in an epistemic
circumstance E, then S semantically ought to accept ‘s’ in E.
First, note that 1d and 2d can come into conflict. 13 Assume that Anna speaks
Swedish, in which ‘häxor existerar’ means
WITCHES EXIST,
it is not the case that
witches exist, yet Anna has sufficient undefeated evidence that they do. In this case,
1d will entail that it is not the case that Anna ought to accept ‘häxor existerar’
whereas 2d will entail that S ought to accept it. Someone who holds that truth is all
that matters for sentence acceptance (belief) would accept 1d but not 2d, whereas
someone who holds that evidence is all that matters for sentence acceptance (belief),
would accept 2d but not 1d. Yet neither of them need be confused about the concept
13
It is worth noting that Gibbard distinguishes between subjective and objective oughts: what
you objectively ought to do is what you ought to do in light of everything that is the case,
what you subjectively ought to do is constrained in some way by your state of information.
He claims that semantic oughts are subjective oughts, which suggests that he would
ultimately reject the implication from 1a to 1b. He can thereby avoid the inconsistency
between the two sets of analytic entailments. But he cannot thereby avoid the objections
presented above, since those turn on the thought that one might sensibly reject either
entailment without displaying conceptual confusion.
20 MEANING.
Their dispute concerns whether one ought to accept a sentence (believe
something) if and only if it is true, or whether one ought to accept a sentence (believe
something) only if one has sufficient undefeated evidence that it is true. It is possible
for someone to endorse either position without thereby displaying confusion about the
concept MEANING.
Perhaps Gibbard would argue that the dispute between the advocate of 1d and
the advocate of 2d is a substantive normative dispute. He might suggest that the
concept
MEANING
is normative in the sense that to grasp the concept
MEANING
one
must accept that meaning ascriptions have some normative implications—one ought
to accept either 1d, 2d, or similar—though it does not matter which normative
implications one takes semantic ascriptions to have. But this too seems to be
mistaken. For, many semantic anti-normativists will be inclined to deny that meaning
ascriptions have any normative entailments. If this denial is sensible, then MEANING is
not a thin normative concept. However, since I count myself among the semantic
anti-normativists, I am not in a position to assess, in an impartial manner, whether our
reasons for denying meaning ascriptions have any normative entailments are
sensible.14
What positive reasons does Gibbard give for accepting the normativity
thesis?15 The chief reason is that the normativity of the concept
MEANING
explains
14
See e.g. Glüer 1999, Glüer-Pagin and Wikforss 2009a, Hattiangadi 2006, 2009, 2007,
Wikforss 2001. For a comprehensive review of the issue, see Glüer-Pagin and Wikforss
2009b.
15
Gibbard (2012: 16) distinguishes between a weak normativity thesis, according to which
meaning ascriptions have normative entailments and a strong normativity thesis, according to
which the meaning of an expression is defined by the pattern of oughts it entails. Though
Gibbard purports to defend both theses, it is only the weak normativity thesis that is relevant
to the present discussion.
21 why certain basic oughts follow from semantic ascriptions ‘invariably.’ (Gibbard
2012:16) But this rings hollow in light of the foregoing discussion. As we have seen
in the cases above, there are sensible grounds to question whether semantic
ascriptions really do entail basic oughts invariably.
Another reason Gibbard provides in support of the normativity thesis has to do
with the hard problem of intentionality. Attempts to analyze
MEANING
in naturalistic
terms have failed, he says, and this can be explained by the hypothesis that MEANING
is normative. (Gibbard 2012: 16-17) But it would be premature to conclude that
MEANING
is normative on this basis. Arguably, phenomenal concepts resist reductive
analysis, but it does not follow that they are normative.
The third reason Gibbard provides is that viewing the concept
MEANING
as
normative promises a satisfactory expressivist resolution to the hard problem of
intentionality. Whether this is a good reason to accept that
MEANING
is a normative
concept depends on whether expressivism does offer a satisfactory resolution to the
hard problem of intentionality. And this remains to be seen.
To assess whether expressivism offers a satisfactory resolution to the hard
problem of intentionality, it will help to have a picture of how the expressivist
solution to the hard problem of morality goes. The expressivist in the moral domain
claims that moral statements do not describe, but express moral judgments, and that
moral judgments are not straightforwardly factual beliefs. In Gibbard’s view, moral
judgments are a special kind of contingency plan. Moral expressivism is attractive to
those who are tempted by nihilism or naturalism. If moral statements express plans,
then there is no need to postulate any normative facts out there in the world to
correspond to them. This gets naturalists of the hook, for if there is no need to
22 postulate any normative facts, there are no normative facts crying out for naturalistic,
reductive explanation. As expressivists argue, there is no need to explain what moral
facts consist in. All that is needed for a fully naturalistic explanation of morality is a
naturalistic explanation of ‘representation and goal-persuit’ (Gibbard 2013:237). In
other words, all that is needed for a fully naturalistic explanation of morality is a
naturalistic explanation of intentionality.
This style of explanation clearly breaks down when applied to the intentional
domain. First of all, the semantic expressivist’s starting point is the semantic thesis
that the concept
MEANING
is normative, and hence that semantic statements do not
describe, but express plans. This thesis appears to commit the expressivist at least to
the existence of one semantic fact—that semantic statements do not describe, but
express plans. For, if this statement itself expressed a plan, it is difficult to see what
bearing the expression of this plan would have on the hard problem of intentionality.
(See Hattiangadi, forthcoming)
Second, in the moral case, we have an explanation for why there is no need to
give a naturalistic, reductive explanation of what goodness consists in—because we
can give a semantics for moral language that does not commit us to normative facts or
properties. But the parallel move in the case of intentionality does not work. The
expressivist can hardly explain why there is no need for a naturalistic explanation of
what meaning consists in by arguing that if the expressivist semantics for meaning
ascriptions is true, then we do not need to postulate any semantic facts or properties.
If semantic ascriptions express plans, as the expressivist claims, then the property of
expressing a plan is a semantic property of semantic ascriptions. This property cries
out for reductive naturalistic explanation.
23 Third, in the moral domain, expressivism is presented as hospitable to
naturalism because it is assumed that it is possible to give a naturalistic explanation of
intentionality. But when applied to intentionality, the expressivist turns out to be
saying that we need not naturalistically explain intentionality, and that all that is
needed is a naturalistic explanation of intentionality—which is blatantly inconsistent.
So, expressivism does not offer a route to a satisfactory solution to the hard problem
of intentionality (See Hattiangadi 2015).
Finally, it is worth noting that for the same reason that the rule following
hypothesis was irrelevant to the hard problem of intentionality, the view that the
concept of meaning is normative is also in a sense irrelevant to this problem.16 As we
noted previously, the hard problem of intentionality has to do with the naturalistic
explanation of an arbitrary representation, including normative concepts, judgments
and statements. The claim that semantic concepts are normative is a hypothesis of
first-order semantic theorizing about semantic concepts, on a par with semantic
analyses of vague expressions, conditionals, or gradable adjectives. Of course, any
reductive explanation of intentionality will have to explain whatever properties are
postulated by an ideal and complete semantic theory. But that semantic theory does
not directly figure in an answer to the hard problem of intentionality.
4 INTENTIONALITY AS A SOURCE OF NORMATIVITY
Let us suppose that meanings and contents are determined by constitutive rules or
normative requirements of some kind. Are these rules or requirements sources of
normativity? Several proponents of the normativity of intentionality have suggested
16
Thanks to John Broome for pointing this out.
24 that they are. For instance, Boghossian claims that what follows from the fact that you
mean green by ‘green’ is a host of normative truths, such as that you ought to apply
‘green’ to something if and only if it is green (Boghossian 1989). Similarly,
Wedgwood does not merely hold that subjects who grasp the concept of the
conditional are disposed accept arguments by modus ponens; he claims further that
they ought to do so. And Ginsborg claims not only that someone who means addition
by ‘plus’ makes a certain pattern of normative judgments, she claims that these
judgments are those one ought to make (see also Whiting 2009). These philosophers
seem to be saying that rules (requirements or principles) that determine meaning and
content are a source of normativity; the semantic facts engender or determine some
normative facts.17 It is because some course of action is required by meaning or
content constitutive rules that one ought to, or has a normative reason to carry out that
course of action.
Following Broome (2013), we can distinguish between a strong and a weak
version of the claim that some norm R is a source of normativity:
Strong Normativity: Necessarily, if R requires that you φ, then you ought to φ
because R requires that you φ.
Weak Normativity: Necessarily, if R requires that you φ, then the fact that R
requires you to φ gives you a normative reason (possibly pro tanto) to φ.
17
What Kathrin Glüer Pagin and Åsa Wikforss call CE normativism CITE
25 Are meaning or content constitutive rules either strongly or weakly normative? First
consider a rule that tells you that you ought to respond with the sum of any two
numbers, when asked. Of course, in many cases, responding with the sum is what you
ought to do, but often when you ought to respond with the sum, it is not because
responding with the sum is required by the rule that constitutes what you mean.
Sometimes, you ought to respond with the sum because you will benefit—if you are
sitting an examination in mathematics, for instance. At other times, you ought to
respond with the sum because this is what you morally ought to do. And sometimes
you ought to respond with the sum because of a desire to communicate what you
believe. But it is not necessarily the case that you ought to respond with the sum
whenever you are asked. Consider again the situation in which you discover that if
you answer with the sum, thousands will die. Is it true that you ought to respond with
the sum in this possible situation simply because this rule requires it? Surely not!
Indeed, it is implausible that meaning constitutive rules such as these are even
weakly normative. The fact that what you mean by ‘plus’ requires that you respond
with the sum does not necessarily give you a pro tanto reason to do so, as if that
reason hung in the balance against the loss of life (Jarvis Thompson 2008). Of course,
you often do have a reason to answer with the sum, as noted above. Such reasons
could of course be present when you discover that if you answer with the sum,
thousands will die. For instance, if at the moment of your discovery, you happen to be
sitting a mathematics exam, you might have to weigh your own benefit against the
lives of others. But in this case, it is not because you are semantically required to
answer with the sum that you have a reason to do so; it is because doing so will
benefit you. And in the absence of any such further reasons to answer with the sum,
26 the mere fact that answering with the sum is required by the meaning of ‘plus’ does
not itself give you any reason to respond with the sum.
It might be suggested once again that there are distinct spheres of normativity:
moral, prudential, epistemic, semantic, and so forth, and that distinct spheres of
normativity come into conflict in the case where you are faced with a choice between
answering with the sum of two numbers and allowing thousands to die, or not
answering with the sum and letting them live: in this situation, what you semantically
ought to do is respond with the sum, and what you morally ought to do is keep quiet.
Even if what you ought to do all things considered is in this case identical with what
you morally ought to do, it does not follow that it is not the case that you semantically
ought to respond with the sum.
However, this point is orthogonal to the present issue. We have assumed, for
the sake of argument, that meaning and content are constituted by rules or normative
requirements of some kind. We have assumed for instance that if you mean addition
by ‘plus’ then the semantics of your language requires that you respond with the sum
of any two numbers when asked. If what you semantically ought to do is just what
you ought to do in view of the relevant semantic requirements, it is trivial that you
semantically ought to do what is semantically required of you. Similarly, it is trivially
true that you legally ought to do what the law requires of you, that you rationally
ought to do what rationality requires of you, and that you morally ought to do what
morality requires of you. But even if we grant all that, there is a further question
whether these systems of norms are sources of normativity, whether you ought to or
have reason to do what these systems of norms require of you. To be reminded that
what you semantically ought to do is what is semantically required of you does not
address this further question.
27 Are norms of content more plausible candidates for sources of normativity?
Consider Wedgwood’s suggestion that if you grasp the concept of ‘if’ you ought to
accept inferences by modus ponens. Now, it is no doubt often true that you ought to
accept inferences by modus ponens. Sometimes, this is because you will benefit by
doing so, and sometimes it is because others will benefit by your doing so. But even if
we suppose that it is necessarily true that you ought to accept inferences by modus
ponens, this is not because you are required to do so by the rule constitutive of the
concept that you grasp.
To see why, consider someone who grasps the McGee conditional rather than
the conditional—call him McGee*. Now suppose, for the sake of argument, that it is
necessarily true that one ought to accept arguments by modus ponens. If this is a
necessary truth, then even if McGee* grasps the McGee conditional, it will be
irrational for him to reject inferences by modus ponens. The fact that McGee* ought
to accept inferences by modus ponens cannot be explained by the requirement
constitutive of the concept that he grasps, for that requirement tells him to reject some
arguments by modus ponens. Perhaps what explains why McGee* ought to accept
inferences by modus ponens is that modus ponens is a requirement of rationality.
Indeed, one of the reasons Wedgwood gives for the normativity thesis is that the
concept of rationality is normative for reasoning (Wedgwood CITE). But if it is
rationality that is normative, then the best explanation of why one ought to accept
inferences by modus ponens is that it is required by rationality, and not by conceptual
requirements. Whatever the explanation why McGee* ought to accept inferences by
modus ponens, it cannot be that accepting inferences by modus ponens is required by
the content of the concept that he grasps. A similar objection can be made against the
view that such rules are weakly normative. Even if it is supposed that necessarily,
28 McGee* has a reason to accept inferences by modus ponens, this cannot be explained
by the requirement of the concept that he grasps.
What bearing might all this have on the hard problem of intentionality? It has
been suggested that if intentionality is a source of normativity, then there will be no
satisfactory solution to the hard problem of intentionality until there is a satisfactory
solution to the hard problem of normativity. It has also been suggested that there can
be no reductive naturalistic solution to the hard problem of normativity because
normative properties are irreducible to the natural properties, and that since
intentionality is a source of normativity, there can be no reductive naturalistic solution
to the hard problem of intentionality either (Boghossian 1989, Wedgwood 2007). All
this is true so far as it goes. But since intentionality is not a source of normativity,
there are no implications for the hard problem of intentionality.
5 NORMATIVITY AND THE REDUCTION OF INTENTIONALITY
Finally, consider the view that the semantic facts are in some sense reducible to the
normative facts (together with the natural facts). In what sense reducible? First, one
might hold that the semantic truths can be reductively explained in terms of the
normative and natural truths. 18 Second, one might hold that the semantic facts
supervene on the normative and natural facts. Of course, one might hold both that the
semantic truths can be reductively explained in terms of the normative and the natural
18
For instance, Wedgwood claims that no account of intentionality can be given that does not
mention some normative facts or properties (CITE). This suggests that normativity figures
essentially in the reductive explanation of intentionality. See also Brandom CITE.
29 truths, and that the former supervene on the latter.19 Indeed, if normativity is essential
to the reductive explanation of intentionality, this could be a reason to hold that
intentionality supervenes on the normative and natural facts.
It is important to recognize that none of these normativist views is of much
interest if intentionality can be reductively explained in naturalistic terms alone. For
then, any normative statements added to a reductive explanation of intentionality
would inessential to the adequacy of the explanation. So, we can assume for the sake
of argument that intentionality cannot be reductively explained in purely naturalistic
terms. This is widely regarded as one of the chief lessons to be learned from Kripke’s
(1982) skeptical argument, particularly his arguments to dispositionalism. Our first
question then is whether appeal to some additional normative truths makes the
reductive explanation of intentionality possible.20
There are broadly speaking two approaches to the reductive explanation of
intentionality: holistic and atomistic. The holistic approach, associated with Lewis
(CITE) and Davidson (CITE), for instance, starts with the totality of relevant nonsemantic information and then assigns beliefs, desires, and meanings in accordance
with certain constraints on interpretation, such as most notably, some kind of
19
There is a third option: one might hold that the semantic truths can be reductively explained
in terms of the normative and natural truths, but that the semantic facts supervene on the
natural facts alone. We have already had occasion to consider one version of this third view:
Gibbard holds that semantic concepts are normative, but that they pick out natural properties.
A Cornell realist about semantic normativity might accept the same view (CITE).
20
It is almost universally accepted that the normative truths supervene on the non-normative
truths. Since supervenience is a transitive relation, if the semantic truths supervene on the
normative truths, does it follow that they supervene on the natural truths? The answer is no,
because the non-normative truths on which the normative truths supervene are not identical to
the natural truths. Indeed, the non-normative truths on which the normative truths supervene
might well be semantic or intentional truths, such as truths concerning the meaning of an
utterance of ‘I promise…’ or the contents of a subject’s intrinsic desires.
30 principle of charity. The atomistic approach, associated with Fodor (CITE), Dretske
(CITE), and Millikan (CITE), for instance, starts by giving a reductive explanation of
mental representations and then attempts to build a reductive explanation of beliefs,
desires and the meanings of sentences on this foundation. The contents of mental
representations are reductively explained in terms of causal co-variation. The simple
idea is that if a mental representation X causally co-varies with the presence of Fs,
then X picks out, or represents the Fs. It is well-known that this simple view faces the
disjunction problem. Suppose that horses typically cause the mental representation
‘horse’ to be tokened,21 but that sometimes, such as when a horsey looking cow is
viewed at a distance across a misty meadow, a cow causes ‘horse’ to be tokened. It
follows from the simple causal co-variation thesis that ‘horse’ refers to either horses
or horsey looking cows. To solve this problem, it is necessary to identify some
property of the disposition to token ‘horse’ in the presence horses that is not shared
with the disposition to token ‘horse’ in the presence of cows, and which determines
that the first disposition is content-determining, while the second disposition is not
(Boghossian 1989). This suggests a role for normativity in the reductive explanation
of content: perhaps it is a normative property that determines whether a disposition is
content-determining.
Ruth Millikan seems to endorse a view along these lines, appealing to what
she refers to as a normative notion of biological function (Millikan 1984). She argues
that a disposition to token a mental representation in response to a stimulus
determines that the stimulus in question is the content of that representation if
representing that stimulus is a biological function of the organism. For example, the
21
Fodor (CITE) uses ‘to token’ as a verb for what happens when a mental representation is
instantiated in the mind in the right kind of way (i.e., in a perception or a belief).
31 image on a frog’s retina triggered by the presence of flies represents flies because
representing flies is the biological function of the image: representing flies
contributed to the reproductive success of the frog’s ancestors and continues to
contribute to its reproductive success. Similarly, it is a biological function of the heart
to pump blood because having a heart that pumps blood contributed to the
reproductive success of organisms with hearts in the past, and continues to contribute
to the reproductive success of organisms with hearts. If it the biological function of
‘horse’ is to pick out the horses, then ‘horse’ refers to horses.
What does this have to do with normativity? Millikan suggests that biological
functions are normative in the sense that they have to do with what how things are
supposed to or ought to behave. The heart ought to pump blood; if it fails to do so, it
malfunctions. Moreover, it is good for an organism with a heart if its heart pumps
blood, bad for it when the heart ceases to do so. Similarly, if the mental representation
‘horse’ ought to represent horses, if it is good for the organism that tokens ‘horse’ that
it represents horses, then horses are what ‘horse’ represents for that organism. And if
‘horse’ ought to represent horses, then this is a normative fact about that
representation, determined by a normative property of the disposition to token ‘horse’
in the presence of horses, namely the property of being good for the organism that
possesses it.
The trouble is that ‘the heart ought to pump blood’, despite containing the
word ‘ought’ does not express a genuinely normative statement. What this sentence
expresses is that the heart pumps blood at all possible worlds ranked highest in the
context, in this case, the ordering imposed by the functioning of the organism in
question (Cf. Kratzer CITE). Roughly glossed, the statement ‘the heart ought to pump
blood’ says that the heart’s pumping blood is a necessary condition for the
32 functioning of the organism. And this is really not a normative claim. For it is
possible to coherently judge that the heart’s pumping blood is a necessary condition
for the functioning of the organism while judging that it ought not to be the case, in
the normative sense, that the heart pumps blood. Compare: MRSA bacteria are
resistant to antibiotics because they produce penicillinase; the biological function of
producing penicillinase is to protect the bacteria from the harmful effects of
penicillin. But the statement ‘MRSA bacteria ought to produce penicillinase’ is not a
normative statement. What it is says is that producing penicillinase is necessary for
the functioning of the MRSA bacteria. And it is perfectly coherent to judge that
producing penicillinase is a necessary condition for the ideal functioning of MRSA
bacteria while simultaneously judging in the genuinely normative sense of ‘ought’
that it ought not to be the case that the MRSA bacteria produce penicillinase. Since
Millikan’s theory does not offer a genuinely normative reductive explanation of
intentionality, we can set it aside here.
A genuinely normative reductive explanation of intentionality (that is
atomistic) would specify some genuinely normative properties of the meaningconstituting dispositions. For instance, one might take a leaf out of Wedgwood’s book
and say that one’s rational dispositions determine the content of ‘if’. Alternatively,
one might say that the content of ‘horse’ is determined in part by the fact that the
disposition to token ‘horse’ in the presence of horses is correct, that is, a disposition
to do what one ought to do. Thus, one might explain what makes it the case that one
grasps the concept of the conditional in terms of one’s rational dispositions, or what
makes it the case that ‘horse’ picks out the horses in terms of one’s correct
dispositions.
33 This type of proposal faces a dilemma. First of all, recall the distinction
between what semantic, rational, prudential and all things considered normative
truths. What one semantically ought to do, for instance, is what one ought to do in
light of what one means, or what one ought to do to satisfy the requirements of the
concept one grasps. What one rationally ought to do is what one ought to do in light
of the requirements of rationality. And what one ought to do all things considered is
what one ought to do in light of all relevant facts. The dilemma faced by the
normativist proposal under consideration goes as follows. If semantic normative
truths are included in the reductive explanation, it renders that explanation circular. If,
on the other hand, only non-semantic normative truths are added to the explanation,
reductive explanation fails.
First, suppose that semantic normative truths are included in the explanation,
such as that one semantically ought to token ‘horse’ in the presence of horses, or that
one semantically ought to accept inferences by modus ponens. To say that one
semantically ought to token ‘horse’ in the presence of horses is just another way of
saying that one ought to do this in light of what one means by ‘horse’. To say that one
semantically ought to accept arguments by modus ponens is to say that one ought to
do this in light of the content of the concept that one grasps. However, if semantic
terms or their equivalents occur in the reductive explanation of intentionality, that
renders the explanation circular. What we wanted to know was whether the semantic
truths could be explained in terms of the non-semantic truths, not whether they could
be explained in terms of some semantic truths.
On the other hand, suppose that we include rational normative truths in the
reductive explanation, or truths about what one ought to do all things considered. In
either case, reductive explanation of intentionality fails. Suppose that accepting
34 arguments by modus ponens is a requirement of rationality. If rationality is a source
of normativity, then it is necessarily true that one ought to accept arguments by modus
ponens. This fact would not explain why someone like McGee* grasps the concept
that he does. Similarly, suppose that your mental representation ‘horse’ picks out the
horses, but that there is a possible situation in which thousands will die unless you
token ‘horse’ in the presence of a horsey looking cow. What you ought to do all
things considered is token ‘horse’ in the presence of the horsey looking cow, but this
does not explain what makes it the case that ‘horse’ picks out all and only the horses.
So, it seems that an atomistic reduction of intentionality is not helped by the addition
of normative truths, unless they are semantically normative truths, in which case the
reductive explanation that results is circular.
Let us now turn to a consideration of the holistic approach to the reduction of
intentionality, such as the view of David Lewis. According to Lewis, reductive
explanation of intentionality is possible only if there is an a priori entailment from the
physical truths, P to the semantic truths, S. If there is an a priori entailment from P to
S, then it is possible for an ideally rational being, who knows P to deduce S a priori,
that is, without appeal to any further empirical information. If there is an a priori
entailment from P to S, an ideal interpreter who knows all of the truths about Karl as a
physical system, should able to deduce the semantic truths about an arbitrary subject,
such as Karl. The radical interpreter can accomplish this feat by appeal to a priori
constraints on interpretation—such as a principle of charity—derived from our
concepts of belief, desire and meaning, as defined by folk psychology (Lewis 1974).
Lewis’ proposal faces several difficulties, the best known of which is the socalled ‘permutation problem’. According to one version of Lewis’ view, called
‘global descriptivism’, the radical interpreter proceeds to assign meanings to the
35 expressions of Karl’s language in two stages, as follows. First, she (somehow)
identifies the totality of sentences that Karl holds true, or would hold true under ideal
conditions. This set of sentences constitutes Karl’s global theory, T. Next, applying
the principle of charity, she assigns meanings to the sentences of T in such a way as
to render true as many sentences of that theory as is plausibly possible. The meanings
of lexical expressions are meant to simply fall out of the assignment of truth-values
and truth-conditions to sentences.
As many philosophers have noticed, however, it is possible to permute the
meanings assigned by an interpretation to the lexical expressions of the language in
such a way that they cancel each other out, so that at the level of sentences, the truthvalues remain the same (Field CITE, Putnam CITE, Quine CITE, Williams CITE).
The permutation problem for the global descriptivist goes as follows: from any
interpretation, x, it is possible to construct a competing interpretation y that is
equivalent to x in the assignment of truth values to sentences of T, but which assigns
bizarre extensions to lexical items. For example, suppose that Karl holds true the
sentence s =‘all emeralds are green.’ Let x be an interpretation that assigns the set of
emeralds to ‘emeralds’ and the set of green things to ‘green’, thereby rendering s true.
We can easily construct an interpretation y that assigns the set of grue things to
‘green’ and the set of emerires to ‘emeralds’, where something is grue iff it is
observed before t and found to be green, or is blue otherwise, and where something is
an emerire iff it is observed before t and found to be an emerald, or is a sapphire
otherwise. Though y is a bizarre interpretation, it too renders s true (Davidson CITE).
Lewis’ solution to the permutation problem is to invoke naturalness. Not all
properties are created equal, according to Lewis: the property of being green is more
natural than the property of being grue. And this, he claims, has implications for
36 metasemantics. All else being equal, an interpretation of Karl which assigns the belief
that all emeralds are green is better than an interpretation which assigns the belief that
all emeralds are grue because the property of being green is more natural than the
property of being grue. However, as Williams has recently shown, the combination of
charity and naturalness do not give rise to the desired result. It is possible to construct
bizarre interpretations that are at least as natural and charitable as the intended
interpretation (Williams CITE. See also Hawthorne CITE).
Now, Lewis’ proposal is intended to be fully reductive and fully naturalistic;
he is not a normativist. But we can nevertheless ask whether the appeal to normative
truths can in any way help resolve the permutation problem. Indeed, some have
thought that normativity is crucial to Davidson’s account of intentionality, which
bears some relation to Lewis’, though Davidson’s aim is not to give a fully reductive
explanation of intentionality (Jackman CITE, Engel CITE).22 So, we can consider the
prospects of a kind of hybrid of Lewis’ view and the normativist view attributed to
Davidson.
Is it possible to avoid the permutation problem by adding normativity to the
explanation? Perhaps we should add to Lewis’ story that the radical interpreter ought
to assign truths to the sentences of T, at least so far as is plausibly possible. Or
perhaps we should say that the justification for the principle of charity lies in the
normative truth that people ought to accept the truth. But even if these are genuinely
normative truths, it does not help in any way to resolve the permutation problem. For,
that problem arises after the principle of charity has been applied.
22
Like Lewis, Davidson holds that radical interpretation is possible. However, Davidson
gives up on the hope of a fully reductive explanation no intentionality (Davidson 1990). Note,
moreover, that it is controversial whether Davidson truly is a normativist. Cf. Glüer CITE,
Davidson CITE.
37 Or perhaps we should say that the principles of rationality that figure in the
radical interpretation of belief and desire are normative. Lewis sometimes suggests
that decision theory is a refinement of folk psychology (Lewis 1974). Perhaps what
Lewis should say is that it is normative decision theory that refines folk psychology, a
theory that characterizes fundamental principles of rationality, rather than descriptive
decision theory, which purports to describe human behavior. Then he might say that
the belief that p is the state that occupies the normative belief-role; it is rationalized
by evidence and in turn rationalizes behavior. So, if the radical interpreter is presumed
to know which beliefs are rational in light of the subject’s evidence, and which beliefs
and desires rationalize the subject’s behavior, she can thereby determine what the
subject means.
But this would not help to resolve the permutation problem, or rather, that
version of the problem that arises for assignments of belief and desire. Suppose that
Karl observes a number of green emeralds, and becomes disposed to utter sentence s,
that interpretation x assigns to Karl the belief that all emeralds are green, while
interpretation y assigns the belief that all emerires are grue. One might try to argue
that whereas interpretation x rationalizes Karl’s belief, interpretation y does not,
because ‘green’ and ‘emerald’ are projectible predicates, while ‘grue’ and ‘bleen’ are
not (Goodman CITE, Lewis CITE, Weatherson CITE). However, this will not do. As
Davidson pointed out, it is the relation between predicates that makes them instance
confirmable, as Davidson showed (1980). Though ‘all emeralds are grue’ may not be
instance confirmable, the pair of laws ‘all emerires are grue’ and ‘all sapheralds are
bleen’ are jointly instance confirmable. Just as an observation of a green emerald
confirms the hypothesis that all emeralds are green, an observation of a grue emerire
confirms the hypothesis that all emerires are grue (Davidson 1980:226).
38 Thus it seems that normativity does not figure essentially in the reductive
explanation of intentionality. What about the view that the semantic facts supervene
on the normative and natural facts? Often, the argument for supervenience turns on
the possibility of reductive explanation. If intentionality cannot be reductively
explained in terms of the normative and natural truths, the supervenience claim cannot
be defended on this basis. Of course, it is a familiar view that reductive explanation is
not needed for supervenience. For instance, so-called Type B naturalists about
consciousness hold that consciousness cannot be reductively explained, but that it
supervenes on the natural facts nonetheless. Similarly, one might argue that though
normativity does not figure in the reductive explanation of intentionality, the semantic
truths supervene on the normative and natural truths nonetheless. The trouble is that it
is difficult to see what is gained by adding normative truths to the supervenience base
if those normative truths do not figure essentially in the reductive explanation. A type
B materialist about intentionality accepts that intentionality cannot be reductively
explained in purely naturalistic terms, and holds that the intentionality supervenes on
the natural nonetheless, in the service of ontological parsimony. If normative truths do
not figure in the reductive explanation of intentionality, adding normative facts to the
the supervenience base would appear to be an ontological extravagance.
6 CONCLUSION In sum, we have seen normativity really has very little bearing on the hard problem of
intentionality, in any of the four senses given above. The view that meaning involves
rule following or a normative judgment of some kind is untenable, and in any case,
has no bearing on the hard problem of intentionality. The same goes for the
hypothesis that the concept of meaning is normative. In contrast, the view that
39 intentionality is a source of normativity would have a bearing on the hard problem of
intentionality if it were true, but since it is not true, it really has no bearing on the hard
problem after all. Finally, nothing is be gained by including normative truths in the
reductive explanation of intentionality, and hence we have little reason to suppose that
intentionality supervenes on any normative facts or properties.
7 WORKS CITED
Boghossian, P. 1989. ‘The Rule-Following Considerations.’ Mind, 98:507-549.
____. 2005. ‘Rules, Meaning and Intention – Discussion.’ Philosophical Studies,
124:185-197.
Brandom, R. 1994. Making it Explicit. Cambridge MA: Harvard University Press.
_____. 2000. Articulating Reasons. Cambridge MA: Harvard University Press.
Broome, J. 2013. Rationality Through Reasoning. Oxford: Oxford University Press.
Chalmers, D. 1996. The Conscious Mind. Oxford: Oxford University Press.
Clifford, W. K. 1877. The ethics of belief. Contemporary Review. Reprinted in T.
Madigan, ed. The ethics of belief and other essays. Amherst, MA: Prometheus,
1999: 70-96.
Davidson, D. 1990. The structure and content of truth. Journal of Philosophy,
87(6):279-328.
Gibbard, A. 1994. ‘Meaning and Normativity.’ Philosophical Issues, 5:95-115.
_____. 1996. ‘Thought, Norms, and Discursive Practice.’ Philosophy and
Phenomenological Research, 56: 699-717.
40 _____. 2003. ‘Thoughts and norms.’ Philosophical Issues, 13: 83-98.
_____. 2013. Meaning and Normativity. Oxford: Oxford University Press.
Ginsborg, H. 2011. ‘Primitive Normativity and Skepticism about Rules.’ The Journal
of Philosophy: 108 (5):227-254.
_____, 2012. ‘Meaning, Understanding and Normativity,’ Aristotelian Society
Supplementary Volume 86 (1): 127-146.
Glüer, K. 1999. ‘Sense and Prescriptivity,’ Acta Analytica, 14: 111-128.
Glüer, K. & Pagin, P. 1999. ‘Rules of Meaning and Pracitcal Reasoning.’ Synthese,
117: 207-227.
Glüer, I. And Wikforss, Å. 2009. ‘Against Content Normativity.’ Mind, 118: 31-70.
_____. 2010. ‘The Normativity of Meaning and Content.’ The Stanford
Encyclopaedia of Philosophy (Winter 2010 edition), Edward N. Zalta (ed.),
URL
=
http://plato.stanford.edu/archives/win2010/entries/meaning-
normativity/.
Hattiangadi, 2003. ‘Making it Implicit: Brandom on Rule Following.’ Philosophy and
Phenomenological Research, LXVI: 419-431.
_____. 2006. ‘Is Meaning Normative?’ Mind and Language, 21: 220-240.
_____. 2007. Oughts and Thoughts: Rule-Following and the Normativity of Content.
Oxford: Oxford University Press.
Ichikawa, J. & Jarvis, B. 2013. The Rules of Thought. Oxford: Oxford University
Press.
41 McHugh, C. 2014. Fitting belief. Proceedings of the Aristotelian Society 114(2):167187.
Millikan, Ruth (1984). Language, Thought, and Other Biological Categories.
Cambridge, Mass.: Bradford/MIT.
Pascal, B. 1670. Pensées. W.F. Trotter, trans. London: Dent, 1910.
Peacocke, C. 1992. A Study of Concepts. Cambridge MA: Harvard University Press.
Thompson, J.J. 2008. Normativity. Peru, IL: Open Court.
Wedgwood, R. 2002. ‘The Aim of Belief.’ Philosophical Perspectives, 16: 276-297.
_____. 2007. The Nature of Normativity. Oxford: Oxford University Press.
_____. 2009. ‘The Normativity of the Intentional.’ The Oxford Handbook of the
Philosophy of Mind. Brian P. McLaughlin and A. Bekermann, eds. Oxford:
Clarendon Press, 421-36.
Whiting, D. 2007. ‘The normativity of meaning defended,’ Analysis, 67:133-40.
____. 2009. Is meaning fraught with ought? The Pacific Philosophical Quarterly,
90(4):575-599.
_____. 2015. Review of Meaning and Normativity, by Allan Gibbard. European
Journal of Philosophy. 23(S1):e14-e18.
Wikforss, Å. 2001. ‘Semantic Normativity.’ Philosophical Studies, 102: 203-226.
Williamson, T. 2007. The Philosophy of Philosophy. Oxford: Blackwell.
Wittgenstein, L. 1953. Philosophical Investigations. Oxford: Blackwell.
42